AaronBallman wrote:
> LLVM can handle IR like the following without ever explicitly allocating
> memory for the array:
>
> ```
> @g = global {[1000000000000 x i32], i32} {[1000000000000 x i32]
> zeroinitializer, i32 3}
> ```
>
> Of course, you hit an out-of-disk-space error if you try to compile this to
> an object file.
Thanks! Yeah, this is the "kicking the can down the road" problem I was worried
about. We'll avoid the OOM from Clang, but once LLVM is processing the IR into
an object file, it'll just OOM again.
One thing I'm concerned by is why we should have this for designated
initializers but not array declarations in general. I would expect all of these
to have effectively the same behavior regarding being too big of an array to
support:
```
int array1[0x80000000];
int array2[0x80000000] = { 0 };
int array3[] = { [0x80000000 - 1] = 0 };
```
So I'm basically worried that `-fmax-init-list-elements` is too narrow of an
option and we should be thinking about a more general flag, if any. What do
others think?
https://github.com/llvm/llvm-project/pull/205503
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
